splash lubrication simulation problem
 

Hi, recently i'm trying to simulate the transient gear splash lubrication in a gear box. but the result turns out to be wrong, i would appreciate your help.

my model is a gear box and three rotating shaft with gear on it, as the picture 1.
http://img1.ph.126.net/dnPmJVR1YUbln...4322020915.jpg

one of the gear shaft is shown as picture 2. I defined this domain as rotate frame, the inner boundary is wall and the outter boundar is interface with the domain of gear box. the other two shaft are defined in the same way.

http://img2.ph.126.net/bmHNcYmYHGiIm...6536305967.jpg
after i defined the initial water and air volume fraction, since the max rotate angle velocity is 1667 rev per minite, i set the timestep 0.0005[s] and the total time is 0.5[s].
in my calculating process, i get the water volume fraction contour in CFXpost. in the 1st step,the volume fraction is Ok and shown as picture 3
http://img0.ph.126.net/dycy9hzPmbfOq...5376710424.jpg
but in the 2ed step the volume fraction seems not right and shown as picture 4. it did not perform as an incessancy process

http://img2.ph.126.net/uAWUt9gAdah81...3234092273.jpg
the transient volume fraction did not perform as an incessancy process. i'm wondering why does this happened. acturally i have made some examples before this simulation, such as one gear rotate in a closed space, and the result seems right. so i do not know what is wrong. if you have some ideas about my case, please tell me.
thank you very much

Your images have not come through. Have a look at "Go Advanced" to post images.

How did you choose your time step size? Did you do a sensitivity analysis to determine it? Or use an adaptive time step criteria? Or did you just guess?

Your question is an FAQ: http://www.cfd-online.com/Wiki/Ansys..._inaccurate.3F

But in this case I bet too big time step is one big factor. Very few people seem to bother to set the time step properly and it leads to all sorts of problems.

thank you for your reply. i have already fix the my problem.
as for timestep, i used to read a paper which recommend the timestep in rotating frame transient simulation to be (1/angle velocity), i do not know it is right or wrong, but i change my timestep to 0.0001 since my max volocity is 1000 rev per minite.
In addition, in my simulation, i pay a lot of attention to the volume fraction. i plan to set air and oil phase as inhomogeneous model and mixture model, but i thought that homogeneous model may also make sence, since the interface of air and water seems to share the same volecity. i have checked the CFX help document, and i do not see distinct statement of homogeneous and inhomogeneous model recommentation, since they both seem to be reasonable for me.
so would you please tell me the model choice more specifically, thank you very much

If the phases remain distinct then use homogeneous. If they mix on a sub grid scale length scale the use inhomogeneous.